Agents that do what you don't know how to.

A wrapper answers you. This acts. The Sales agent buys and warms its own sending domains, holds credentials for your CRM and calendar, decides on Tuesday that Thursday's subject line needs changing, and does it without asking. The model is a component. The product is everything around it: the permissions, the tools, the memory of what it already tried, and the judgement about when to stop.

The agent sends from addresses it owns and warms over three weeks precisely so it doesn't. It verifies every address before writing, keeps daily volume inside the limits that keep you out of spam folders, and every message carries a working one-click unsubscribe. Anyone who opts out is never contacted again by any agent on your account, permanently, and no one can reverse it.

You see every message before the first campaign goes out, and you can keep approval on permanently if you want it. Most people turn it off in week two. Either way, the agent shows you what it wrote, what changed since last week, and why, in plain language, not a diff.

That's exactly the gap this exists to close. Every number the agent shows you comes with what normal looks like next to it. 1 to 2 percent is typical for cold outreach, 4 percent is good. You should never have to go and research whether your own results are fine.

You set direction and you can stop anything in one click, at any point, without losing the work in progress. What you hand over is the execution: the hundred small craft decisions a week you'd otherwise be making badly, slowly, at midnight, in between shipping.

Because sending a hundred cold emails a day from the domain your team runs on is how you lose the domain your team runs on. Separate names, bought in your company's name at what they cost us, kept warm and monitored by the agent. You keep them if you leave. It's a good example of the whole idea: a decision you didn't know you had to make, and shouldn't have to.

They land in the same workspace, with the same data and the same permissions model, and they talk to each other. The Support agent knows what Sales promised. The Marketing agent knows which messages got replies. That shared context is the reason this is an operating system rather than a folder full of separate tools.

2,000 credits, and a place in the next batch. One credit is one thing an agent does: one person found, one address checked, one message written, one sent. So 2,000 is roughly 400 people found, written to and followed up. That is a full Starter month, which is $39 after launch. The credits land the day we switch your account on, not the day you pay, because credits you cannot spend are not credits.

You get the $9 back. It is refundable in full at any point before we switch your account on, with no time limit and no explanation required. Email hello@operater.io and it goes back the same day. We would rather refund you than argue with you: a refund costs us a processing fee, and a chargeback costs us far more, so there is no version of this where being difficult pays.

Automation runs the steps you already decided on. You build the sequence, set the rules, and it executes them the same way every time; if a reply does not match a branch you wrote, nothing happens. An agent decides the steps. It reads your product, works out who the buyer is, writes to each one differently, and handles a reply it has never seen before without you adding a rule for it. The practical difference is who does the thinking: with automation the thinking is yours and the tool supplies the typing, which is the wrong way round when you have never done the job.

It is built for exactly that, and the honest answer has a boundary. If you have shipped something and cannot get anyone to look at it, the agent does the part you have never done and gives you the transcript of what it tried, so you learn the job while it runs. If you already have a salesperson and a working process, the gain is small, because the execution layer you would be replacing is already staffed. And if your product needs a long consultative sale into an enterprise, a human closes that, not this. Operater gets you to the conversation; it does not pretend to be the person having it.